WO2016021412A1

WO2016021412A1 - Coding device and method, decoding device and method, and program

Info

Publication number: WO2016021412A1
Application number: PCT/JP2015/070924
Authority: WO
Inventors: 修一郎錦織; 鈴木　志朗
Original assignee: ソニー株式会社
Priority date: 2014-08-06
Filing date: 2015-07-23
Publication date: 2016-02-11
Also published as: EP3608910A1; US10049677B2; US20180322885A1; JP2016038435A; EP3179476A4; US20170270940A1; EP3179476B1; US10510353B2; EP3179476A1; CN106663449B; CN106663449A; EP3608910B1

Abstract

This patent relates to a coding device and method, a decoding device and method, and a program which enable high-quality sound to be obtained even in an environment with small resources. A decomposition unit decomposes a supplied code string to obtain a quantized low-frequency spectrum, a spectral characteristic code, and a quantized extension coefficient. On this occasion, a single quantized extension coefficient or a quantized extension coefficient for each high-frequency band is included in the code string according to the spectral characteristic code. A spectrum inverse quantization unit inversely quantizes the quantized low-frequency spectrum to obtain a low-frequency spectrum, and an extension coefficient inverse quantization unit inversely quantizes the quantized extension coefficient to obtain an extension coefficient. An extended spectrum generation unit generates an extended spectrum on the basis of the low-frequency spectrum and the extension coefficient corresponding to the spectral characteristic code. An IMDCT unit generates a band-extended time-series signal from the low-frequency spectrum and the extended spectrum. This patent is applicable to a decoding device.

Description

Encoding apparatus and method, decoding apparatus and method, and program

The present technology relates to an encoding apparatus and method, a decoding apparatus and method, and a program, and in particular, an encoding apparatus and method, a decoding apparatus and method, and a decoding apparatus that can obtain high-quality sound even in a low resource environment, and Regarding the program.

Conventionally, an encoding technique that incorporates the concept of bandwidth expansion for audio signals is known (see, for example, Patent Document 1 and Patent Document 2).

In such an encoding technique, a time-series signal input as an audio signal is band-divided into a low-frequency component and a high-frequency component, and normal encoding is performed on the low-frequency signal, and the low-frequency signal and The relationship of the high frequency signal, the characteristics of the high frequency signal, and the like are transmitted as additional information.

At the time of decoding, after the low-frequency signal is restored, the low-frequency signal and the additional information are used to generate an extended-band signal, and the low-frequency signal and the extended-band signal are combined, Bandwidth expansion is realized.

More specifically, after the low-frequency signal is restored, the low-frequency signal is divided into a plurality of bands by a band division filter, and the divided low-frequency signal and additional information are used. Thus, an extended band signal is generated. Then, the low-band signal and the extension band signal are synthesized by the band synthesis filter, and a band-expanded time-series signal is obtained.

However, when the band division filter and the band synthesis filter are used in this way, the principle delay from the encoding to the decoding of the signal is increased by the band division and band synthesis filter processing. If it does so, the response speed from the input of an audio | voice signal to an output will fall.

In addition to normal decoding processing, filter processing such as band division and band synthesis using a filter bank is required, which greatly increases the amount of processing and memory usage, and decoding is possible in low-resource environments such as embedded devices. It was difficult to install the device.

Therefore, as a technique for improving such an encoding technique, a technique that enables band extension in the frequency domain has been proposed (see, for example, Patent Document 3).

With this technology, the spectrum obtained by MDCT (Modified Discrete Cosine Transform) at the time of encoding is divided into a low frequency side (baseband) and a high frequency side (extended band), and normal encoding is performed for baseband signals. As a result, the relationship between the spectrum of the baseband and the extension band, the characteristics of the extension band spectrum, and the like are transmitted as additional information.

Also, at the time of decoding, the baseband spectrum and the additional information are used to generate an extension band spectrum, and the baseband spectrum and the extension band spectrum are synthesized to generate a full band spectrum. Further, an IMDCT (Inverse Modulated Discrete Cosine Transform) is performed on the obtained spectrum of the entire band, whereby the spectrum of the entire band is converted into a time series signal (time signal).

Japanese Patent No. 5329714 Japanese Patent No. 5325293 JP 2011-215198 A

However, the value of each frequency bin of the spectrum obtained by MDCT (hereinafter also referred to as MDCT spectrum) is a value in which both the amplitude component and the phase component are woven. Therefore, in the technology that performs band extension in the frequency domain, if the amplitude of the spectrum in the extension band is finely adjusted using the MDCT spectrum at the time of decoding, the phase component of each spectrum and the mutual phase relationship between each spectrum are greatly destroyed. .

In such a case, for example, when the audio signal to be encoded and decoded is a signal such as a noisy musical tone or a human voice, the audio signal does not significantly deteriorate in sound quality.

However, if the audio signal is an audio signal in which energy is concentrated at a specific frequency such as a single musical instrument or sound effect, that is, a signal with high tonality, the energy that should have been concentrated at the specific frequency is decoded. Will spread into the spectrum of the surrounding frequency. If it does so, the audio | voice signal finally obtained by decoding will have noise property, and sound quality on hearing will deteriorate.

As described above, the technology for performing band expansion in the frequency domain does not require band division or band synthesis for a time-series signal, so that voice encoding and decoding can be performed even in a low resource environment without causing a delay. However, there were cases where high-quality sound could not be obtained.

The present technology has been made in view of such a situation, and is capable of obtaining high-quality sound even in a low-resource environment.

The decoding device according to the first aspect of the present technology includes a low-frequency spectrum and a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low-frequency band, or a plurality of components constituting the extension band An acquisition unit that acquires an extension coefficient for each band; a generation unit that generates the extension spectrum based on the single extension coefficient or the extension coefficient for each of the plurality of bands; and the low-frequency spectrum; And a synthesis unit that synthesizes the extended spectrum.

The generation unit can generate the extended spectrum based on the low frequency spectrum and the extension coefficient.

The generation unit can generate the extended spectrum by adjusting the level of the spectrum obtained from the low-frequency spectrum based on the extension coefficient.

When generating the extended spectrum based on the single extension coefficient, the generation unit adjusts the level of the entire extension band of the spectrum based on the extension coefficient, and When generating the extended spectrum based on an extension coefficient, the level of the band of the spectrum can be adjusted based on the extension coefficient of the band.

The generation unit can generate the extended spectrum by adjusting a predetermined noise level based on the extension coefficient.

The value of the low frequency spectrum can be determined by the amplitude component and the phase component of the original time series signal.

The low frequency spectrum can be an MDCT spectrum.

The decoding method or program according to the first aspect of the present technology configures a low-band spectrum and a single extension coefficient for the extension band to obtain an extension spectrum of an extension band different from the low band, or the extension band Obtaining an extension coefficient for each of a plurality of bands, generating the extension spectrum based on the single extension coefficient or the extension coefficient for each of the plurality of bands, and adding the low-frequency spectrum and the extension spectrum to each other. Synthesizing.

In the first aspect of the present technology, a low-frequency spectrum, a single expansion coefficient for the expansion band for obtaining an expansion spectrum of an expansion band different from the low-frequency band, or each of a plurality of bands constituting the expansion band And the extension spectrum is generated based on the single extension coefficient or the extension coefficient for each of the plurality of bands, and the low-frequency spectrum and the extension spectrum are combined.

The encoding device according to the second aspect of the present technology includes a feature amount extraction unit that extracts a feature amount from a spectrum obtained by orthogonally transforming a time-series signal, and a low-frequency region of the spectrum according to the feature amount. A calculation unit for calculating, based on the spectrum, a single extension coefficient for the extension band for obtaining an extension spectrum of a different extension band, or an extension coefficient for each of a plurality of bands constituting the extension band; And a multiplexing unit that multiplexes a low-frequency spectrum that is a low-frequency component and the extension coefficient to generate a code string.

The feature quantity can be information indicating the tonality of the spectrum.

The calculation unit can calculate the single extension coefficient when the tonality of the spectrum is high, and can calculate the extension coefficient for each of the plurality of bands when the tonality of the spectrum is low.

The calculation unit can calculate the ratio between the average amplitude of the extension band of the spectrum and the average amplitude of the low band spectrum as the extension coefficient.

The calculation means can calculate envelope information of the extension band of the spectrum as the extension coefficient when the low-frequency tonality of the spectrum is high and the tonality of the extension band of the spectrum is low.

The value of the spectrum can be determined by the amplitude component and phase component of the time series signal.

The orthogonal transform can be MDCT.

The encoding method or program according to the second aspect of the present technology extracts a feature amount from a spectrum obtained by orthogonally transforming a time-series signal, and expands differently from the low band of the spectrum according to the feature amount. A single extension coefficient for the extension band for obtaining an extension spectrum of a band, or an extension coefficient for each of a plurality of bands constituting the extension band is calculated based on the spectrum, and a low band component of the spectrum is calculated. The method includes a step of multiplexing a band spectrum and the extension coefficient to generate a code string.

In the second aspect of the present technology, a feature amount is extracted from a spectrum obtained by orthogonally transforming a time-series signal, and an extension spectrum having an extension band different from the low band of the spectrum is determined according to the feature amount. A single extension coefficient for the extension band to be obtained or an extension coefficient for each of a plurality of bands constituting the extension band is calculated based on the spectrum, and a low-frequency spectrum that is a low-frequency component of the spectrum; The extension coefficient is multiplexed and a code string is generated.

According to the first and second aspects of the present technology, high-quality sound can be obtained even in a low resource environment.

Note that the effects described here are not necessarily limited, and may be any of the effects described in the present disclosure.

It is a figure which shows the structural example of an encoding apparatus. It is a figure explaining the area | region and boundary of a spectrum. It is a figure explaining a low region return pseudo amplitude spectrum. It is a figure explaining division | segmentation of a high region spectrum. It is a flowchart explaining an encoding process. It is a figure which shows the structural example of a decoding apparatus. It is a flowchart explaining a decoding process. It is a figure explaining a signal with high tonality. It is a figure explaining the average value of a high region pseudo amplitude spectrum. It is a figure explaining level adjustment of an extended spectrum. It is a figure explaining collapse of tonality by level adjustment. It is a figure explaining level adjustment of an extended spectrum. It is a figure which shows the example of a signal with a high low region tonality and a low high region tonality. It is a figure explaining the production | generation of an extended spectrum, and sound quality degradation. It is a figure explaining the production | generation of an envelope coefficient and an extended spectrum. It is a flowchart explaining an encoding process. It is a flowchart explaining a decoding process. It is a figure which shows the structural example of a computer.

Hereinafter, embodiments to which the present technology is applied will be described with reference to the drawings.

<First Embodiment>
<Configuration example of encoding device>
FIG. 1 is a diagram illustrating a configuration example of an embodiment of an encoding device to which the present technology is applied.

The encoding apparatus 11 shown in FIG. 1 includes an MDCT unit 21, a spectrum quantization unit 22, a low frequency feature quantity extraction unit 23, a high frequency feature quantity extraction unit 24, a spectrum characteristic determination unit 25, an extension coefficient calculation unit 26, an extension coefficient quantum. And a multiplexing unit 28.

The MDCT unit 21 is supplied with an input signal that is a time-series signal having a sampling frequency Fs [kHz], for example, as a speech signal to be encoded.

The MDCT unit 21 performs, for example, MDCT as orthogonal transform on the supplied input signal, and obtains a spectrum from the frequency Dc [kHz], which is a DC component, to a frequency Fs / 2 that is half the sampling frequency Fs.

In the following, description will be continued by taking MDCT as an example of orthogonal transform, but the spectrum value obtained by orthogonal transform may be a value in which both amplitude component and phase component are woven. For example, any conversion is not limited to MDCT.

In addition, here, in order to improve the encoding efficiency, what is actually encoded is a component from the frequency Dc of the spectrum obtained by the orthogonal transform to the frequency Fc [kHz] that is sensitive to hearing, The remaining spectrum shall be discarded. That is, it is assumed that the part of the spectrum from the frequency Fc to the frequency Fs / 2 is lost.

Suppose that the band expansion is performed on the decoding side in order to further improve the encoding efficiency.

For example, as shown in FIG. 2, it is assumed that the spectrum obtained by orthogonal transformation in the MDCT unit 21 is divided into a low-frequency spectrum, a high-frequency spectrum, and a loss spectrum. In FIG. 2, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, the component from the frequency Dc, which is the direct current component, to the upper limit frequency Fb [kHz] in the entire spectrum is the low frequency spectrum, and when the input signal is encoded, normal encoding is performed on the low frequency spectrum. Done.

Also, the component from the upper limit frequency Fb to the frequency Fc in the whole spectrum is a high frequency spectrum. When the input signal is encoded, the high-frequency spectrum is not encoded, but at the time of decoding, the low-frequency spectrum and an extension coefficient that is additional information described later are used to generate a pseudo high-frequency spectrum (hereinafter referred to as an extended spectrum). (Also referred to as a spectrum) is generated, and band extension is realized. That is, at the time of decoding, the frequency band from the upper limit frequency Fb to the frequency Fc is set as an extension band that is a target of band extension.

Furthermore, the portion from the frequency Fc to the frequency Fs / 2 in the whole spectrum is regarded as a loss spectrum and is lost.

In the following, the band from the frequency Dc to the upper limit frequency Fb is referred to as a low band, and the band from the upper limit frequency Fb to the frequency Fc is referred to as a high band. Hereinafter, a band from the frequency Fc to the frequency Fs / 2 is referred to as a loss band.

Therefore, in this example, only the low frequency component is encoded with the input signal, and the high frequency component is generated by band expansion at the time of decoding.

Returning to the description of FIG. 1, the MDCT unit 21 performs MDCT on the input signal, and converts the low band spectrum of the spectrum of the entire band obtained as a result of the spectrum quantization unit 22 and the low band feature amount extraction unit 23. And the high frequency spectrum is supplied to the high frequency feature amount extraction unit 24.

The spectrum quantization unit 22 quantizes the low frequency spectrum supplied from the MDCT unit 21 and supplies the quantized low frequency spectrum obtained as a result to the multiplexing unit 28.

The low-frequency feature quantity extraction unit 23 extracts a feature quantity (hereinafter also referred to as a low-frequency spectrum feature quantity) from the low-frequency spectrum supplied from the MDCT unit 21 and supplies the extracted feature quantity to the spectral characteristic determination unit 25 and the low-frequency spectrum. Is supplied to the expansion coefficient calculation unit 26.

The high frequency feature quantity extraction unit 24 extracts a feature quantity (hereinafter also referred to as a high frequency spectrum feature quantity) from the high frequency spectrum supplied from the MDCT unit 21, supplies the feature quantity to the spectral characteristic determination unit 25, and Is supplied to the expansion coefficient calculation unit 26.

Here, the low-frequency spectrum feature value and the high-frequency spectrum feature value will be described.

In order to extract feature quantities such as a low-frequency spectrum feature quantity and a high-frequency spectrum feature quantity from the spectrum obtained by the MDCT section 21, it is necessary to observe the spectrum amplitude characteristics. However, the spectrum obtained by the MDCT unit 21 is, for example, an MDCT spectrum obtained by MDCT, and the MDCT spectrum has a property different from a DFT spectrum obtained by DFT (DiscretecreFourier Transform). The MDCT spectrum is also called an MDCT coefficient.

Specifically, the DFT spectrum includes an amplitude component and a phase component independently of each other. On the other hand, the value of the MDCT spectrum, that is, the value in each frequency bin of the MDCT spectrum is a value in which both the amplitude component and the phase component are woven. That is, the value of the MDCT spectrum is determined by the amplitude component and the phase component of the input signal, and the value of only one of the amplitude component and the phase component cannot be known from the value of the MDCT spectrum.

Therefore, when using the DFT spectrum, it is possible to observe the amplitude of the signal using the amplitude spectrum or the power spectrum, but in the case of the MDCT spectrum, the amplitude of the signal is directly derived from the MDCT spectrum. It is difficult to observe.

Therefore, it is conceivable that IMDCT, which is the inverse transform of MDCT, is performed on the MDCT spectrum, the input signal is once converted back to a time-series signal, and then DFT is performed on the time-series signal for feature quantity extraction.

However, in such a case, it is necessary to further add an IMDCT or DFT processing block to the encoding device 11, and the amount of calculation and the amount of memory used such as ROM (Read Only Memory) or RAM (Random Access Memory) A significant increase is expected. If it does so, it will become difficult to encode an input signal in the low resource environment where calculation resources, such as portable equipment, were restricted.

Therefore, the encoding apparatus 11 according to the present technology, based on the MDCT spectrum to calculate the pseudo amplitude spectrum S _k by the following equation (1), used for feature extraction.

In the equation (1), the pseudo amplitude spectrum S _k shows the pseudo amplitude spectrum corresponding to the k-th frequency bin of the MDCT spectrum, y _k is a value of the MDCT spectrum corresponding to the k-th frequency bin Show. Therefore, in Equation (1), the pseudo amplitude spectrum _Sk is calculated for one frequency bin based on the value of the MDCT spectrum corresponding to three consecutive frequency bins.

The value of the pseudo amplitude spectrum S _k obtained in this way is a similar value to the amplitude spectrum. That is, the value of the pseudo amplitude spectrum S _{k is} a value having a strong correlation with the amplitude spectrum of the DFT spectrum, and therefore the value of the pseudo amplitude spectrum S _k indicates a pseudo amplitude value at each frequency of the MDCT spectrum. be able to.

In the following, the pseudo-amplitude spectrum obtained for the low-frequency spectrum is also referred to as a low-frequency pseudo-amplitude spectrum, and the pseudo-amplitude spectrum obtained for the high-frequency spectrum is particularly referred to as a high-frequency pseudo-amplitude spectrum.

Low range characteristic amount extraction section 23 and the high-range feature extraction unit 24, the low frequency band spectrum and high spectrum of each frequency (frequency bins), and calculates the pseudo amplitude spectrum S _k by equation (1), the resulting It calculates a characteristic quantity from the pseudo amplitude spectrum S _k for each frequency bin.

For example, the low-frequency feature quantity extraction unit 23 and the high-frequency feature quantity extraction unit 24 indicate the high noise characteristics of the spectrum by the calculation of the following equation (2) as the low-frequency spectrum feature value and the high-frequency spectrum feature value. Spectral Flatness (hereinafter also referred to as SF) serving as an index is calculated.

In Equation (2), N indicates the number of target spectra, that is, the number of frequency bins. Further, S _i represents the value of the pseudo amplitude spectrum of the i-th frequency bin.

Therefore, for example, when calculating SF for the high frequency spectrum, the arithmetic average of the pseudo amplitude spectrum S _k calculated for all frequency bins of the high frequency spectrum, relative to the geometric average of the pseudo amplitude spectrum S _k calculated for all frequency bins of the high frequency spectrum The ratio is SF.

The SF calculated in this way indicates the degree of flatness of the spectrum and takes a value in the range of 0.0 to 1.0.

For example, the larger the SF value, that is, the closer the SF value is to 1.0, the smaller and flatter the spectrum is, and the higher the noise characteristics of the spectrum. Conversely, the smaller the SF value, that is, the closer the SF value is to 0.0, the higher the spectral tonality (lower noise).

Although an example in which SF is calculated as a feature amount has been described, any feature amount may be calculated.

For example, other than SF, there is an index indicating the high level of noise in the spectrum, in other words, an index indicating the high level of tonality, so that it depends on the accuracy of the feature amount required by the encoding device 11 and the allowable calculation amount. Thus, another index indicating the high noise property may be calculated as the feature amount.

As an example of a feature quantity different from SF, for example, a spectrum concentration degree D shown in the following equation (3) may be calculated as a low-frequency spectrum feature quantity or a high-frequency spectrum feature quantity.

In Equation (3), N indicates the number of target spectra, that is, the number of frequency bins. S _i indicates the value of the pseudo amplitude spectrum corresponding to the i th frequency bin, and Max (S _i ) indicates the maximum value of the pseudo amplitude spectrum S _i corresponding to each frequency bin.

Therefore, in the example of Expression (3), the ratio of the arithmetic average of the pseudo amplitude spectrum S _{k to} the maximum value of the pseudo amplitude spectrum S _k is the spectrum concentration degree D.

In the MDCT spectrum, the larger the value of the spectrum concentration degree D, the more uneven the distribution of the spectrum and the higher the tonality. On the contrary, the smaller the value of the spectrum concentration degree D, the more flat the distribution of the spectrum and the higher the noise characteristic.

As described above, any feature amount may be calculated, but the description will be continued below assuming that SF is calculated as the feature amount.

Specifically, when the low-frequency feature quantity extraction unit 23 calculates the low-frequency spectrum feature quantity, as shown in FIG. 3, the low-frequency pseudo-amplitude spectrum calculated for the low-frequency spectrum is high with the upper limit frequency Fb as a boundary. The above-described SF is calculated for the low-frequency aliasing pseudo amplitude spectrum obtained by aliasing to the band side.

In FIG. 3, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, the low-frequency pseudo-amplitude spectrum represented by the curve C11 is folded back to the high-frequency side at the position of the upper limit frequency Fb to be the low-frequency aliasing pseudo-amplitude spectrum represented by the curve C12. Therefore, the low frequency pseudo amplitude spectrum and the low frequency aliasing pseudo amplitude spectrum are symmetrical waveforms.

Returning to the description of FIG. 1, the low frequency feature quantity extraction unit 23 calculates the frequency bins of the range from the upper limit frequency Fb to the frequency Fc in the low frequency aliasing pseudo-amplitude spectrum obtained by the aliasing by the equation (2). SF is calculated as a low-frequency spectrum feature value by calculation. In the following, SF calculated as a low-frequency spectrum feature amount is particularly referred to as SFL.

The low frequency feature quantity extraction unit 23 supplies the SFL as the low frequency spectrum feature quantity obtained in this way to the spectrum characteristic determination unit 25, and also uses the low frequency aliasing pseudo amplitude spectrum as the amplitude information as the expansion coefficient calculation unit 26. To supply. At this time, for example, a portion from the upper limit frequency Fb to the frequency Fc in the low frequency aliasing pseudo amplitude spectrum is supplied to the expansion coefficient calculation unit 26.

Further, the high frequency feature quantity extraction unit 24 calculates SF as the high frequency spectrum feature quantity by calculating the equation (2) for each frequency bin of the high frequency pseudo amplitude spectrum obtained from the high frequency spectrum. In the following, SF calculated as a high-frequency spectrum feature is particularly referred to as SFH.

The high frequency feature quantity extraction unit 24 supplies the SFH as the high frequency spectrum feature value obtained in this way to the spectrum characteristic determination unit 25 and also supplies the high frequency pseudo amplitude spectrum to the expansion coefficient calculation unit 26 as amplitude information. Supply.

Based on the low-frequency spectrum feature quantity supplied from the low-frequency feature quantity extraction unit 23 and the high-frequency spectrum feature quantity supplied from the high-frequency feature quantity extraction unit 24, the spectral characteristic determination unit 25 A spectral characteristic code indicating the spectral characteristic of the input signal is generated.

For example, when both SFL, which is a low-frequency spectrum feature quantity, and SFH, which is a high-frequency spectrum feature quantity, are less than a predetermined threshold, the spectrum characteristic code is a code indicating high tonality. In other words, the input signal (MDCT spectrum) has a spectral characteristic that the tonality is high. Here, the value of the spectrum characteristic code indicating high tonality is assumed to be “1”.

Also, when at least one of SFL, which is a low-frequency spectrum feature quantity, and SFH, which is a high-frequency spectrum feature quantity, is greater than or equal to a threshold value, the spectrum characteristic code is a code indicating that it is not high tonality. That is, the input signal has a spectral characteristic that the tonality is not high, in other words, the noise property is high. Here, it is assumed that the value of the spectrum characteristic code indicating that the tonality is not high is “0”.

As described above, when the tonality is high in both the low frequency component and the high frequency component of the MDCT spectrum, the spectral characteristic code is “1”, and the noise property of at least one of the low frequency component and the high frequency component of the MDCT spectrum is set. Is high, the spectral characteristic code is set to “0”.

The spectrum characteristic determination unit 25 supplies the spectrum characteristic code obtained in this way to the extension coefficient calculation unit 26, the extension coefficient quantization unit 27, and the multiplexing unit 28.

The expansion coefficient calculation unit 26 is based on the low-frequency aliasing pseudo-amplitude spectrum from the low-frequency feature quantity extraction unit 23, the high-frequency pseudo-amplitude spectrum from the high-frequency feature quantity extraction unit 24, and the spectrum characteristic code from the spectrum characteristic determination unit 25. The expansion coefficient is calculated and supplied to the expansion coefficient quantization unit 27.

Here, the expansion coefficient is information for performing level adjustment of the high frequency in the frequency domain at the time of decoding, and indicates the ratio of the levels of the high frequency pseudo amplitude spectrum and the low frequency aliasing pseudo amplitude spectrum. In other words, the expansion coefficient indicates the ratio between the average amplitude of the high frequency spectrum and the average amplitude of the low frequency spectrum.

Specifically, when the spectrum characteristic code is “1”, the expansion coefficient calculation unit 26 is a high frequency range, that is, the value of the high frequency pseudo amplitude spectrum of each frequency bin in the band from the upper limit frequency Fb to the frequency Fc. The average value of is calculated. Further, the expansion coefficient calculation unit 26 calculates the average value of the low-frequency aliasing pseudo-amplitude spectrum of each frequency bin in the band from the upper limit frequency Fb to the frequency Fc, and calculates the average value of the high-frequency pseudo-amplitude spectrum as the low frequency A value obtained by dividing by the average value of the aliasing pseudo amplitude spectrum is defined as an expansion coefficient. In this case, one expansion coefficient is obtained for the entire high band, that is, the entire expansion band.

On the other hand, when the spectrum characteristic code is “0”, the expansion coefficient calculation unit 26 divides from the low frequency side to the high frequency side in consideration of human auditory characteristics, for example, as shown in FIG. The high band is divided into a plurality of bands so that the obtained bandwidth becomes wide, and the expansion coefficient is calculated for each band.

In FIG. 4, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, the frequency band of the high band spectrum, that is, the frequency band from the upper limit frequency Fb to the frequency Fc, which is the high band, is divided into five bands, band B1 to band B5. The width of each band obtained by the division is wider as the band is on the frequency Fc side.

The expansion coefficient calculation unit 26 obtains the average value of the high-frequency pseudo-amplitude spectrum value by dividing the average value of the low-frequency aliasing pseudo-amplitude spectrum value for each of the bands B1 to B5 constituting the high frequency band. The obtained value is calculated, and the obtained value is set as the expansion coefficient of each band.

For example, the value obtained by dividing the average value of the high frequency pseudo amplitude spectrum in each frequency bin in the band B1 by the average value of the low frequency aliasing pseudo amplitude spectrum in each frequency bin in the band B1 is the band B1. An expansion coefficient of

Accordingly, the expansion coefficient C _i of the i-th band (region) obtained by dividing the high region is calculated by the following equation (4).

In Equation (4), S _k represents the value of the high frequency pseudo-amplitude spectrum of the k th frequency bin in the i th band, and L _k represents the k th frequency bin in the i th band. The value of the low-frequency aliasing pseudo amplitude spectrum is shown. M represents the number of spectra in the i-th band, that is, the number of frequency bins.

The expansion coefficient quantization unit 27 quantizes the expansion coefficient supplied from the expansion coefficient calculation unit 26 based on the spectral characteristic code supplied from the spectral characteristic determination unit 25, and multiplexes the quantized expansion coefficient obtained as a result. To the conversion unit 28.

For example, when the spectrum characteristic code is “1”, scalar quantization is performed on a single extension coefficient calculated for the entire high frequency band. On the other hand, when the spectral characteristic code is “0”, scalar quantization or vector quantization is performed on a plurality of extension coefficients calculated for each divided band (region) in the high band.

The multiplexing unit 28 multiplexes the quantized low frequency spectrum from the spectrum quantizing unit 22, the spectrum characteristic code from the spectrum characteristic determining unit 25, and the quantized expansion coefficient from the expansion coefficient quantizing unit 27, and obtains the result. The encoded code string is output. At this time, the multiplexing unit 28 performs entropy encoding on the quantized low frequency spectrum and also encodes the quantization extension coefficient.

<Description of encoding process>
Next, the operation of the encoding device 11 will be described.

For example, when an input signal to be encoded is supplied from the outside, the encoding device 11 starts an encoding process and encodes the input signal. Hereinafter, the encoding process performed by the encoding device 11 will be described with reference to the flowchart of FIG.

In step S11, the MDCT unit 21 performs MDCT on the supplied input signal. Then, the MDCT unit 21 supplies the low-frequency part of the MDCT spectrum obtained by MDCT as a low-frequency spectrum to the spectrum quantization unit 22 and the low-frequency feature quantity extracting unit 23, and the high-frequency part of the MDCT spectrum. Is supplied to the high frequency feature quantity extraction unit 24 as a high frequency spectrum.

In step S12, the spectrum quantization unit 22 quantizes the low frequency spectrum supplied from the MDCT unit 21, and supplies the quantized low frequency spectrum obtained as a result to the multiplexing unit 28.

In step S13, the low frequency feature amount extraction unit 23 extracts a low frequency spectrum feature value from the low frequency spectrum supplied from the MDCT unit 21.

For example, the low frequency feature quantity extraction unit 23 calculates the above-described equation (1) for each frequency bin of the low frequency spectrum, and calculates a low frequency pseudo amplitude spectrum.

Also, the low frequency feature quantity extraction unit 23 folds the obtained low frequency pseudo amplitude spectrum to the high frequency side at the upper limit frequency Fb to obtain a low frequency aliasing pseudo amplitude spectrum. At this time, for example, the low-frequency feature quantity extraction unit 23 generates a low-frequency aliasing pseudo-amplitude spectrum by discarding a portion having a frequency higher than the frequency Fc of the aliasing low-frequency pseudo-amplitude spectrum.

Then, the low frequency feature quantity extraction unit 23 calculates the above-described formula (2) for each frequency bin of the low frequency aliasing pseudo amplitude spectrum, and calculates SFL as the low frequency spectrum feature quantity.

The low frequency feature quantity extraction unit 23 supplies the SFL as the calculated low frequency spectrum feature value to the spectrum characteristic determination unit 25 and supplies the low frequency aliasing pseudo amplitude spectrum to the expansion coefficient calculation unit 26.

In step S14, the high frequency feature quantity extraction unit 24 extracts a high frequency spectrum feature value from the high frequency spectrum supplied from the MDCT unit 21.

For example, the high frequency feature quantity extraction unit 24 calculates the above-described equation (1) for each frequency bin of the high frequency spectrum to calculate the high frequency pseudo amplitude spectrum, and also calculates the formula ( The calculation of 2) is performed, and SFH is calculated as a high-frequency spectrum feature amount.

The high frequency feature quantity extraction unit 24 supplies SFH as the calculated high frequency spectrum feature value to the spectrum characteristic determination unit 25 and also supplies the high frequency pseudo amplitude spectrum to the expansion coefficient calculation unit 26.

In step S15, the spectrum characteristic determination unit 25, based on the low frequency spectrum feature quantity supplied from the low frequency feature quantity extraction unit 23 and the high frequency spectrum feature quantity supplied from the high frequency feature quantity extraction unit 24, A spectral characteristic code indicating the spectral characteristic is generated.

Specifically, the spectrum characteristic determination unit 25 selects a spectrum characteristic code having a value of “1” when both the SFL that is the low-frequency spectrum feature quantity and the SFH that is the high-frequency spectrum feature quantity are less than the threshold. Generate.

On the other hand, when at least one of SFL, which is a low-frequency spectrum feature quantity, and SFH, which is a high-frequency spectrum feature quantity, is greater than or equal to a threshold value, the spectrum characteristic determination unit 25 has a value of “0”. A certain spectral characteristic code is generated.

The spectrum characteristic determination unit 25 supplies the generated spectrum characteristic code to the extension coefficient calculation unit 26, the extension coefficient quantization unit 27, and the multiplexing unit 28.

In step S <b> 16, the expansion coefficient calculation unit 26 and the expansion coefficient quantization unit 27 determine whether or not the spectral characteristic indicates high tonality based on the spectral characteristic code supplied from the spectral characteristic determination unit 25. .

For example, when the value of the spectrum characteristic code is “1”, it is determined that the spectrum characteristic indicates high tonality.

If it is determined in step S16 that the tonality is high, the process proceeds to step S17.

In step S <b> 17, the expansion coefficient calculation unit 26 applies to the entire high frequency band based on the low frequency aliasing pseudo amplitude spectrum from the low frequency feature value extraction unit 23 and the high frequency pseudo amplitude spectrum from the high frequency feature value extraction unit 24. Thus, a single (one) expansion coefficient is calculated and supplied to the expansion coefficient quantization unit 27.

That is, the expansion coefficient calculation unit 26 calculates the average value of the high frequency pseudo amplitude spectrum in each frequency bin and the average value of the low frequency aliasing pseudo amplitude spectrum in each frequency bin for the band from the upper limit frequency Fb to the frequency Fc. Divide by value to calculate expansion factor.

After the expansion coefficient is calculated, the process proceeds to step S19.

On the other hand, if it is determined in step S16 that it does not indicate high tonality, the process proceeds to step S18.

In step S18, the expansion coefficient calculation unit 26 divides the high frequency band based on the low frequency aliasing pseudo amplitude spectrum from the low frequency feature value extracting unit 23 and the high frequency pseudo amplitude spectrum from the high frequency feature value extracting unit 24. An expansion coefficient is calculated for each band, and is supplied to the expansion coefficient quantization unit 27.

That is, for example, the expansion coefficient calculation unit 26 divides the entire high frequency band into five bands B1 to B5 as shown in FIG. 4 and performs the calculation of the above-described formula (4) for each band, The expansion coefficient is calculated for each. In this case, one extension coefficient is calculated for each of the bands B1 to B5.

When the expansion coefficient is calculated in step S17 or step S18, in step S19, the expansion coefficient quantization unit 27 quantizes the expansion coefficient supplied from the expansion coefficient calculation unit 26, and the quantized expansion coefficient obtained as a result thereof. Is supplied to the multiplexing unit 28.

In step S20, the multiplexing unit 28 multiplexes the quantized low frequency spectrum from the spectrum quantizing unit 22, the spectrum characteristic code from the spectrum characteristic determining unit 25, and the quantized expansion coefficient from the expansion coefficient quantizing unit 27. Generate a code string. At this time, the multiplexing unit 28 encodes the quantized low frequency spectrum and the quantized extension coefficient, and then multiplexes the encoded quantized low frequency spectrum and the quantized extended coefficient, and the spectrum characteristic code.

The multiplexing unit 28 outputs the code string obtained by multiplexing, and the encoding process ends.

As described above, the encoding device 11 determines the spectral characteristics of the input signal based on the low-frequency spectrum feature value and the high-frequency spectrum feature value. And the encoding apparatus 11 calculates a different expansion coefficient according to a spectrum characteristic as an expansion coefficient for adjusting the level of a high region in a frequency domain at the time of decoding.

Thereby, it is possible to adjust the high frequency level in the frequency domain using the expansion coefficient at the time of decoding, and it is possible to realize the high frequency level adjustment according to the spectrum characteristics. Therefore, high-quality sound can be obtained even in a low resource environment without increasing the principle delay.

That is, since the level adjustment of the high frequency can be performed in the frequency domain, the time delay due to the band expansion at the time of decoding is reduced, and an increase in resources on the decoding side is also suppressed. In addition, high-frequency level adjustment can be performed according to the spectral characteristics, so it is possible to suppress deterioration of the audible sound quality even with high tonality signals or low tonality signals, and to obtain higher-quality sound. Will be able to.

<Configuration example of decoding device>
Next, a decoding device that decodes the code string output from the encoding device 11 will be described.

FIG. 6 is a diagram illustrating a configuration example of an embodiment of a decoding device to which the present technology is applied.

6 includes a decomposition unit 91, a spectrum inverse quantization unit 92, an extended coefficient inverse quantization unit 93, an extended spectrum generation unit 94, and an IMDCT unit 95.

The code sequence output from the multiplexing unit 28 of the encoding device 11 is supplied to the decomposition unit 91. The decomposition unit 91 decomposes the supplied code string and obtains a quantized low frequency spectrum, a spectrum characteristic code, and a quantization extension coefficient from the code string. The decomposition unit 91 also decodes the quantized low frequency spectrum and the quantized expansion coefficient.

The decomposing unit 91 supplies the quantized low-frequency spectrum obtained from the code string to the spectrum inverse quantizing unit 92, and the spectrum characteristic code obtained from the code string is expanded to an extension coefficient inverse quantizing unit 93 and an expanded spectrum generating unit. 94. In addition, the decomposition unit 91 supplies the quantization extension coefficient obtained from the code string to the extension coefficient inverse quantization unit 93.

The spectrum inverse quantization unit 92 inversely quantizes the quantized low frequency spectrum supplied from the decomposition unit 91 and supplies the obtained low frequency spectrum to the extended spectrum generation unit 94 and the IMDCT unit 95. The expansion coefficient inverse quantization unit 93 dequantizes the quantized expansion coefficient supplied from the decomposition unit 91 based on the spectrum characteristic code supplied from the decomposition unit 91, and converts the obtained expansion coefficient into the extended spectrum generation unit 94. To supply.

Based on the spectral characteristic code supplied from the decomposition unit 91, the extended spectrum generation unit 94 uses the extended coefficient supplied from the extended coefficient inverse quantization unit 93 and the low frequency spectrum supplied from the spectrum inverse quantization unit 92. An extended spectrum is generated and supplied to the IMDCT unit 95.

The IMDCT unit 95 sets the low frequency spectrum supplied from the spectrum inverse quantization unit 92 as a low frequency spectrum and the extended spectrum supplied from the extended spectrum generation unit 94 as a high frequency (extended band) spectrum. Combine (synthesize) band spectrum and extended spectrum. Further, the IMDCT unit 95 performs IMDCT orthogonal transformation on the spectrum obtained by combining the low-frequency spectrum and the extended spectrum, and outputs the resulting time-series signal as a speech signal obtained by decoding. To do.

<Description of decryption processing>
Next, the operation of the decoding device 81 will be described.

When the code sequence is supplied, the decoding device 81 starts a decoding process, decodes the code sequence, and outputs an audio signal. Hereinafter, the decoding processing by the decoding device 81 will be described with reference to the flowchart of FIG.

In step S51, the decomposition unit 91 decomposes the supplied code string, and obtains a quantized low frequency spectrum, a spectrum characteristic code, and a quantization extension coefficient from the code string.

The decomposition unit 91 supplies the obtained quantized low-frequency spectrum to the spectrum inverse quantization unit 92, supplies the spectrum characteristic code to the extension coefficient inverse quantization unit 93 and the extension spectrum generation unit 94, and performs quantization extension. The coefficient is supplied to the extended coefficient inverse quantization unit 93. In more detail, the decomposing unit 91 decodes the quantized low frequency spectrum and the quantized extension coefficient, and the decoded quantized low frequency spectrum and the quantized extended coefficient are converted into the spectrum inverse quantizing unit 92 and the extended The coefficient is supplied to the coefficient inverse quantization unit 93.

In step S52, the spectrum inverse quantization unit 92 inversely quantizes the quantized low frequency spectrum supplied from the decomposition unit 91, and supplies the obtained low frequency spectrum to the extended spectrum generation unit 94 and the IMDCT unit 95.

In step S53, the extended coefficient inverse quantization unit 93 and the extended spectrum generation unit 94 determine whether or not the spectral characteristic shows high tonality based on the spectral characteristic code supplied from the decomposition unit 91.

For example, when the value of the spectrum characteristic code is “1”, it is determined that the spectrum characteristic indicates high tonality. In this case, since the code string includes a quantization extension coefficient for obtaining one (single) extension coefficient calculated for the entire high band, the expansion coefficient inverse quantization is performed from the decomposition unit 91. The unit 93 is supplied with one quantization expansion coefficient.

Conversely, when the value of the spectral characteristic code is “0”, it is determined that the spectral characteristic does not indicate high tonality, that is, indicates high noise characteristics. In this case, since the code string includes quantization extension coefficients for obtaining each extension coefficient calculated for each of a plurality of bands constituting the high band, the decomposition unit 91 to the extension coefficient inverse quantization unit 93 Are supplied with quantization expansion coefficients corresponding to the number of high-frequency divided bands.

If it is determined in step S53 that the tonality is high, in step S54, the expansion coefficient inverse quantization unit 93 inversely quantizes the single quantization expansion coefficient supplied from the decomposition unit 91. The extended coefficient is supplied to the extended spectrum generation unit 94.

In step S55, the extended spectrum generation unit 94 generates an extended spectrum based on the single extension coefficient supplied from the extension coefficient inverse quantization unit 93 and the low frequency spectrum supplied from the spectrum inverse quantization unit 92. , Supplied to the IMDCT section 95.

Specifically, the extended spectrum generation unit 94 folds the low band spectrum to the high band side with the upper limit frequency Fb as the boundary in the same manner as the example described with reference to FIG. , A seed spectrum for obtaining an extended spectrum.

The extended spectrum generation unit 94 multiplies the entire obtained seed spectrum, that is, the value of the seed spectrum in each frequency bin by a single extension coefficient to obtain an extended spectrum. That is, the level of the seed spectrum is adjusted to the level of the original high-frequency spectrum before encoding by the extension coefficient, and is set as the extended spectrum.

The extended spectrum obtained in this way is the high-frequency spectrum of the original input signal estimated from the low-frequency spectrum obtained by decoding and the expansion coefficient.

When the extended spectrum is obtained, the process proceeds to step S58.

On the other hand, if it is determined in step S53 that the spectral characteristic does not indicate high tonality, that is, indicates high noise, the process proceeds to step S56.

In step S56, the extension coefficient inverse quantization unit 93 inversely quantizes the quantization extension coefficient for each of a plurality of bands constituting the high frequency supplied from the decomposition unit 91, and uses the obtained extension coefficient as the extension spectrum generation unit 94. To supply. Thereby, for example, the expansion coefficient of each band (area) of the bands B1 to B5 shown in FIG. 4 is obtained.

In step S57, the extended spectrum generation unit 94 generates an extended spectrum based on the extension coefficient of each band supplied from the extension coefficient inverse quantization unit 93 and the low frequency spectrum supplied from the spectrum inverse quantization unit 92. , Supplied to the IMDCT section 95.

Specifically, the extended spectrum generation unit 94 generates a seed spectrum by performing the same process as in step S55, and for each band (region) of the obtained seed spectrum, the extension coefficient of those bands To obtain an extended spectrum.

For example, when the high band is divided into five bands B1 to B5 as shown in FIG. 4, the value of the seed spectrum in the band B1 portion of the seed spectrum, more specifically, each frequency bin in the band B1. Is multiplied by the expansion coefficient of band B1 to generate the band B1 portion of the extended spectrum. Similarly, for the other bands B2 to B5, those bands of the seed spectrum are multiplied by the extension coefficient of each band, and each band portion of the extended spectrum is generated.

When the extended spectrum is obtained, the process proceeds to step S58.

In step S55 and step S57, the example in which the low-frequency spectrum is turned back to the high-frequency side is used as the seed spectrum. However, the seed spectrum may be generated in any way. For example, a spectrum obtained by duplicating (copying) a part of a part of the frequency band of the low-frequency spectrum and pasting it on the high frequency may be used as the seed spectrum.

When an extended spectrum is generated in step S55 or step S57, in step S58, the IMDCT unit 95 is based on the low frequency spectrum supplied from the spectrum inverse quantization unit 92 and the extended spectrum supplied from the extended spectrum generation unit 94. To generate a time series signal.

That is, the IMDCT unit 95 combines (synthesizes) the low-frequency spectrum and the extended spectrum to generate a spectrum having all the band components of the low-frequency and high-frequency (extended bandwidth). IMDCT is performed to obtain a time series signal. Thereby, a time series signal to which a high frequency component is added by band expansion is obtained.

The IMDCT section 95 outputs the time-series signal obtained in this way as an audio signal obtained by decoding, and the decoding process ends.

As described above, the decoding device 81 obtains the extension coefficient corresponding to the spectrum characteristic by decoding and inverse quantization, the obtained extension coefficient, and the seed spectrum obtained by folding the low-frequency spectrum to the high-frequency side. Generate an extended spectrum from

In this way, by using the expansion coefficient according to the spectral characteristics, the level of the seed spectrum that is a high frequency component is adjusted, and by setting the expanded spectrum, the high frequency level can be adjusted in the frequency domain, High frequency level adjustment according to the spectrum characteristics can be realized.

This makes it possible to obtain high-quality sound even in a low-resource environment without increasing the principle delay. That is, by performing level adjustment in the frequency domain, it is possible to reduce delay time due to band expansion during decoding and to suppress an increase in resources. In addition, it is possible to suppress the deterioration of sound quality due to the band expansion for a signal having a high tonality or a signal having a low tonality, and to obtain a higher quality sound.

<About generation of extended spectrum>
Here, generation of an extended spectrum by the extended spectrum generation unit 94 of the decoding device 81 will be described in more detail.

As described above, the extended spectrum generation unit 94 distinguishes whether the original signal before encoding is a signal with high tonality or a normal signal with high noise characteristics based on the spectrum characteristic code. Is generated.

For example, as shown in FIG. 8, a signal having a high tonality and a normal signal having a high noise characteristic have different spectral outlines. In FIG. 8, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In FIG. 8, a curve C21 represents a high noise signal, that is, a normal signal spectrum, and a curve C22 represents a high tonal signal spectrum.

The highly noisy signal represented by curve C21 has no protruding portion in the entire frequency band, and the spectrum waveform has a gentle mountain shape. In other words, a signal with high noise characteristics does not have a portion where energy is concentrated.

On the other hand, the signal with high tonality represented by the curve C22 has energy concentrated on a specific frequency, and the waveform of the portion is like a sharply pointed mountain. That is, the waveform of the spectrum of the signal with high tonality is not a gentle waveform because the frequency portion where energy is concentrated protrudes.

In addition, when generating an extended spectrum, it can be obtained from the low-frequency spectrum, such as the low-frequency spectrum folded at the upper limit frequency Fb, or the low-frequency spectrum partially copied and pasted at the high frequency. The spectrum is used as the seed spectrum. Then, this kind of spectrum is level-adjusted by the expansion coefficient, that is, the amplitude is adjusted to obtain an expanded spectrum.

Here, in a normal signal with high noise characteristics, the phase relationship between adjacent ones in each spectrum is not so important for hearing, and the amplitude level is important. Therefore, when adjusting the level of the seed spectrum, it is desirable to perform level adjustment in fine units so that the level (amplitude) of the seed spectrum is as close as possible to the level of the high-frequency spectrum of the original signal before encoding.

That is, for example, as shown in FIG. 9, it is assumed that the high frequency band is divided into four bands at the time of encoding, and the expansion coefficient is calculated for each band. In FIG. 9, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, the frequency band of the high-frequency spectrum, that is, the frequency band from the upper limit frequency Fb to the frequency Fc, which is a high frequency, is divided into four bands (areas) B11 to B14. The width of each band obtained by the division is wider as the band is on the frequency Fc side.

In such a case, in the encoding of the input signal, the average value of the high-frequency pseudo-amplitude spectrum in each of the bands B11 to B14 is calculated. In this example, each of the straight lines L11 to L14 represents the average value of the high frequency pseudo-amplitude spectrum in each of the bands B11 to B14, that is, the average amplitude of the high frequency spectrum.

Further, a value obtained by dividing the average value of the high frequency pseudo amplitude spectrum obtained for each band by the average value of the low frequency aliasing pseudo amplitude spectrum of the same band is stored in the code string as an extension coefficient, and is decoded. 81.

Then, in the decoding device 81, as shown in FIG. 10, the level of the seed spectrum obtained from the low frequency spectrum is adjusted by the extension coefficient. In FIG. 10, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency. In FIG. 10, the same reference numerals are given to the portions corresponding to those in FIG. 9, and the description thereof will be omitted as appropriate.

In FIG. 10, a curve C31 represents a low-frequency spectrum obtained by decoding a code string, and a curve C32 represents a seed spectrum obtained from the low-frequency spectrum.

In this example, the low-frequency spectrum represented by the curve C31 is folded back to the high-frequency side at the upper limit frequency Fb to obtain the seed spectrum represented by the curve C32.

Each of the bands B11 to B14 of such a seed spectrum is multiplied by each of the expansion coefficients calculated for each band. As a result, the level of the seed spectrum is adjusted so that the average amplitude of each band of the seed spectrum, more specifically, the average amplitude of each band approaches the average amplitude of the high-frequency spectrum of the original signal as indicated by an arrow in the figure. Adjustment is made in each of the bands B11 to B14.

However, when the low-frequency spectrum is a signal with high tonality, if the seed spectrum is multiplied by a different expansion coefficient for each band, the level of each band of the expanded spectrum, that is, the average amplitude, is the original high-frequency spectrum before encoding. However, the phase relationship of the spectrum is greatly broken in each band.

Then, for example, the tonality of the extended spectrum is impaired as shown in FIG. In FIG. 11, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, curve C41 represents the MDCT spectrum of the input signal to be encoded, and curve C42 combines the low-frequency spectrum and extended spectrum generated when decoding the input signal to be encoded. Represents the spectrum obtained. Therefore, in this example, in the spectrum represented by the curve C42, a portion from the frequency Dc to the upper limit frequency Fb is a low-frequency spectrum, and a portion from the upper limit frequency Fb to the frequency Fc is an extended spectrum.

In this example, the original input signal is a signal with high tonality in both low and high frequencies. When decoding such an input signal, if the level of the seed spectrum is adjusted with a different expansion coefficient for each high-frequency band, the phase relationship of the spectrum is significantly disrupted, as shown by the curve C42, and the tonality of the extended band is impaired. It will be.

In the spectrum represented by the curve C42, the waveform of the high frequency region, that is, the extended spectrum is broken, and the tonality that the original MDCT spectrum had has been damaged. In particular, the waveform tends to collapse at the boundary between the high-frequency divided bands, and the tonality is easily lost.

The seed spectrum obtained by folding the low-frequency spectrum maintains the phase relationship of the spectrum in the state as it is, that is, before the level adjustment by the expansion coefficient, and thus the tonality is also maintained.

However, unless the level (amplitude) of the seed spectrum is adjusted, the amplitude level of the high frequency spectrum of the original input signal cannot be reflected in the extended spectrum. As a result, the volume of the high frequency band, that is, the expansion band is different from that of the original high frequency band, so that appropriate band expansion cannot be realized. In other words, it becomes impossible to obtain higher quality sound.

Therefore, in this technology, both the retention of the tonality in the extended spectrum and the reflection of the amplitude level are realized by adjusting the level of the seed spectrum in the minimum unit for a signal with high tonality.

Specifically, at the time of encoding, the expansion coefficient calculation unit 26 divides the average value of the high frequency pseudo amplitude spectrum in the entire high frequency band (extension frequency band) by the average value of the low frequency aliasing pseudo amplitude spectrum in the entire high frequency band. Then, a single expansion coefficient is calculated for the expansion band.

Further, at the time of decoding, the extended spectrum generation unit 94 multiplies the entire seed spectrum by a single extension coefficient to obtain an extended spectrum. That is, the level of the seed spectrum is adjusted by using the entire extended band (high band) as a unit to obtain an extended spectrum.

Thus, by performing level adjustment in units of the extension band, for example, as shown in FIG. 12, while maintaining the tonality of the input signal, the overall amplitude level of the high frequency range of the extended spectrum is also the high frequency of the original input signal. It can be close to the amplitude level. In FIG. 12, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In FIG. 12, curves C51 to C53 represent the MDCT spectrum of the original input signal, the low-frequency spectrum obtained by inverse quantization at the time of decoding, and the seed spectrum, respectively.

In this example, the MDCT spectrum represented by the curve C51 has a low-frequency part and a high-frequency part, that is, a low-frequency spectrum and a high-frequency spectrum, each of which has energy concentrated at a specific frequency. The signal is high. In the MDCT spectrum represented by the curve C51, the average amplitude of the low frequency spectrum is larger than the average amplitude of the high frequency spectrum.

For such a high-frequency spectrum of the MDCT spectrum, at the time of encoding, an average value of the high-frequency pseudo-amplitude spectrum is obtained for the entire band of the high-frequency spectrum, and a single expansion coefficient is calculated. In FIG. 12, a straight line L21 represents the average value of the high-frequency pseudo-amplitude spectrum in the high frequency band (extended band), that is, the average amplitude of the high frequency spectrum.

At the time of decoding, the low-frequency spectrum represented by the curve C52 is folded back to be a seed spectrum represented by the curve C53, and the level of the seed spectrum is adjusted by the expansion coefficient as represented by the arrow in the figure. And an extended spectrum.

At that time, the average amplitude of the entire high frequency range of the extended spectrum is made closer to the average value of the high frequency pseudo amplitude spectrum represented by the straight line L21 by a single expansion coefficient. Thereby, since the level of each frequency of the seed spectrum is adjusted by the same amount, the amplitude level can be adjusted appropriately without breaking the phase relationship, that is, while maintaining the tonality. As a result, higher quality sound can be obtained.

Further, if the extension coefficient is single, the amount of additional information necessary for band extension stored in the code string output from the encoding device 11 can be reduced, so that the low-frequency spectrum is correspondingly reduced. It is possible to allocate the amount of information to the quantization of the sound, and to improve the overall sound quality.

<Second Embodiment>
<Generation of extended spectrum by random noise>
By the way, when the low-frequency tonality of the input signal is high, the high-frequency tonality is usually high. Therefore, in the encoding process described above, when both the low-frequency spectrum feature value and the high-frequency spectrum feature value are less than the threshold value, the input signal to be encoded has a spectral characteristic that the tonality is high. It was said.

However, although the frequency is not high, for example, as shown in FIG. 13, there is an input signal having a spectral characteristic that the tonality of the low-frequency spectrum is high and the tonality of the high-frequency spectrum is low. In FIG. 13, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In FIG. 13, curve C61 represents the MDCT spectrum of the input signal to be encoded. In particular, in this MDCT spectrum, a portion from the frequency Dc to the upper limit frequency Fb is a low frequency spectrum, and a portion from the upper limit frequency Fb to the frequency Fc is a high frequency spectrum.

For example, in the low frequency spectrum, there is a portion where energy is concentrated at a specific frequency, which is a signal with high tonality. On the other hand, the high frequency spectrum does not have a portion where energy is concentrated at a specific frequency and is a signal with low tonality, that is, a signal with high noise.

Suppose that low-frequency tonality is high in this way, but high-frequency tonality encodes a low input signal and performs band expansion at the time of decoding. In such a case, when a seed spectrum is generated by folding or partial duplication of a low-frequency spectrum and an extended spectrum is generated from the seed spectrum, for example, as shown in FIG. May appear strongly. In FIG. 14, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In this example, a curve C71 represents a low-frequency spectrum obtained by inverse quantization of the quantized low-frequency spectrum, and a curve C72 represents an extended spectrum.

In this example, the high-frequency spectrum of the original time-series signal has a low tonality, but the low-frequency spectrum has a high tonality, so the extended spectrum obtained by folding the low-frequency spectrum and adjusting the level using the expansion coefficient Has a high tonality. That is, a characteristic different from the characteristic of the original signal appears in the high band due to the band expansion.

When high tonalities that were not originally possessed in the high frequency band appear in this way, the time series signal (audio signal) obtained by the decoding process may cause a sense of incongruity, for example, metal sounds may be mixed. turn into.

Therefore, when the low-frequency spectrum tonality is high and the high-frequency spectrum tonality is low, the extended spectrum is obtained using random noise as shown in FIG. 15, for example, without using the low-frequency spectrum aliasing as a seed spectrum. You may make it produce | generate. In FIG. 15, the vertical axis indicates the spectrum value, that is, the level, and the horizontal axis indicates the frequency.

In FIG. 15, curves C81 to C83 represent the MDCT spectrum, the low-frequency spectrum obtained by dequantizing the quantized low-frequency spectrum, and the extended spectrum, respectively.

In this example, the high region of the MDCT spectrum is divided into three bands, band B31 to band B33, and the higher the frequency, the wider the bandwidth. When the high frequency band is divided into bands B31 to B33, an envelope coefficient is calculated as envelope information indicating the envelope of the band for each band during encoding. For example, the envelope coefficient is an average value of the high-frequency pseudo-amplitude spectrum of each frequency bin in the calculation target band.

In FIG. 15, each of the straight lines L31 to L33 indicates the envelope coefficient calculated for each of the bands B31 to B33.

The envelope coefficient is expansion coefficient information for adjusting the level of random noise as a noise signal when generating an extended spectrum. Here, the envelope coefficient is distinguished from the expansion coefficient calculated from the low-frequency aliasing pseudo-amplitude spectrum and the high-frequency pseudo-amplitude spectrum. Therefore, it will be referred to as an envelope coefficient. It should be noted that the number of high frequency divisions at the time of calculating the envelope coefficient may be the same as or different from the number of high frequency divisions at the time of calculating the expansion coefficient.

When the envelope coefficient is calculated, the envelope coefficient is quantized and encoded, and multiplexed with the quantized low frequency spectrum and spectrum characteristic code to generate a code string.

Also, on the decoding side that receives the supply of the code string, an extended spectrum is generated using the envelope coefficient acquired from the code string and random noise.

That is, at the time of decoding, a random number normalized to a value in the range of -1.0 to 1.0 is generated for each frequency bin of the bands B31 to B33 which are the extension bands, and a noise signal composed of a random number for each frequency bin is generated. Random noise. Then, the random spectrum is multiplied by an envelope coefficient to obtain an extended spectrum.

Since the extended spectrum obtained in this way is generated from random noise obtained by normalizing random numbers, as shown in the curve C83, the energy is not concentrated on a specific frequency and a spectrum with high noise characteristics is obtained. It has become. In addition, since the extended spectrum is obtained by adjusting the level of random noise using the envelope coefficient, the envelope is close to the high-frequency envelope of the original MDCT spectrum.

Therefore, the time-series signal obtained by decoding has a high low-frequency spectrum tonality and a low high-frequency spectrum tonality, similar to the encoded original input signal.

<Description of encoding process>
Next, an encoding process performed by the encoding device 11 when the envelope coefficient described above is generated will be described.

Hereinafter, the encoding process performed by the encoding device 11 will be described with reference to the flowchart of FIG. Note that the processing from step S91 to step S94 is the same as the processing from step S11 to step S14 in FIG.

In step S95, the spectrum characteristic determination unit 25, based on the low frequency spectrum feature quantity supplied from the low frequency feature quantity extraction unit 23 and the high frequency spectrum feature quantity supplied from the high frequency feature quantity extraction unit 24, A spectral characteristic code indicating the spectral characteristic is generated.

Specifically, the spectrum characteristic determination unit 25 selects a spectrum characteristic code having a value of “1” when both the SFL that is the low-frequency spectrum feature quantity and the SFH that is the high-frequency spectrum feature quantity are less than the threshold. Generate. The spectrum characteristic code “1” indicates that both low and high frequencies of the input signal (MDCT spectrum) have high tonality as a spectrum characteristic.

Further, the spectrum characteristic determining unit 25 generates a spectrum characteristic code having a value of “2” when the SFL that is the low-frequency spectrum feature quantity is less than the threshold value and the SFH that is the high-frequency spectrum feature quantity is equal to or greater than the threshold value. To do. The spectral characteristic code “2” has a high tonality in the low frequency (low frequency spectrum) of the input signal, and a low tonality, that is, a high noise property in the high frequency (high frequency spectrum) of the input signal. It is shown that.

Further, the spectrum characteristic determination unit 25 generates a spectrum characteristic code having a value of “0” when the SFL that is the low-frequency spectrum feature amount is equal to or greater than the threshold value. The spectral characteristic code “0” indicates that the input signal has low tonality as a spectral characteristic.

In step S96, the expansion coefficient calculation unit 26 and the expansion coefficient quantization unit 27 indicate tonalities in which both low-frequency and high-frequency spectral characteristics are high, based on the spectral characteristic code supplied from the spectral characteristic determination unit 25. It is determined whether or not there is.

For example, when the value of the spectrum characteristic code is “1”, it is determined that the low-frequency and high-frequency spectral characteristics indicate high tonality.

If it is determined in step S96 that the low-frequency and high-frequency spectral characteristics indicate high tonality, the process proceeds to step S97.

In step S97, the expansion coefficient calculation unit 26 applies to the entire high frequency band based on the low frequency aliasing pseudo amplitude spectrum from the low frequency feature value extracting unit 23 and the high frequency pseudo amplitude spectrum from the high frequency feature value extracting unit 24. Thus, a single expansion coefficient is calculated and supplied to the expansion coefficient quantization unit 27.

In step S97, processing similar to that in step S17 in FIG. 5 is performed. When the expansion coefficient is calculated in step S97, the process proceeds to step S101.

If it is determined in step S96 that the low-frequency and high-frequency spectral characteristics do not indicate high tonality, the process proceeds to step S98.

In step S98, based on the spectrum characteristic code, the extension coefficient calculation unit 26 and the extension coefficient quantization unit 27 indicate tonalities with high spectral characteristics in the low band and low tonal characteristics with high spectral characteristics. Determine whether or not.

For example, when the value of the spectrum characteristic code is “2”, it is determined that the low band spectral characteristic indicates high tonality and the high band spectral characteristic indicates low tonality.

If it is determined in step S98 that the low frequency spectrum characteristic indicates high tonality and the high frequency spectral characteristic indicates low tonality, the process proceeds to step S99.

In step S99, the expansion coefficient calculation unit 26 calculates an envelope coefficient for each divided high frequency band based on the high frequency pseudo-amplitude spectrum from the high frequency characteristic amount extraction unit 24, and the expansion coefficient quantization unit 27 To supply.

That is, for example, the expansion coefficient calculation unit 26 divides the entire high frequency band into three bands B31 to B33 as shown in FIG. 15, and calculates the average value of the high frequency pseudo amplitude spectrum of the frequency bins in each band. It is calculated as an envelope coefficient of those bands.

After the envelope coefficient is calculated, the process proceeds to step S101.

On the other hand, if it is not determined in step S98 that the low band spectral characteristic indicates high tonality and the high band spectral characteristic indicates low tonality, the process proceeds to step S100.

In step S100, the expansion coefficient calculation unit 26 divides the high frequency band based on the low frequency aliasing pseudo amplitude spectrum from the low frequency feature value extraction unit 23 and the high frequency pseudo amplitude spectrum from the high frequency feature value extraction unit 24. An expansion coefficient is calculated for each band, and is supplied to the expansion coefficient quantization unit 27. In step S100, the same process as in step S18 of FIG. 5 is performed. When the expansion coefficient is calculated in step S100, the process proceeds to step S101.

When the expansion coefficient is calculated in step S97 or step S100, or when the envelope coefficient is calculated in step S99, in step S101, the expansion coefficient quantization unit 27 supplies the expansion coefficient or envelope supplied from the expansion coefficient calculation unit 26. Quantize the coefficients.

That is, when the expansion coefficient quantization unit 27 performs the process of step S97 or step S100 and is supplied with the expansion coefficient, the expansion coefficient quantization unit 27 quantizes the expansion coefficient, and the quantized expansion coefficient obtained as a result is sent to the multiplexing unit 28. Supply. Further, when the process of step S99 is performed and the envelope coefficient is supplied, the extended coefficient quantization unit 27 quantizes the envelope coefficient and supplies the resulting quantized envelope coefficient to the multiplexing unit 28. At this time, for example, scalar quantization or vector quantization is performed on the expansion coefficient or the envelope coefficient.

In step S <b> 102, the multiplexing unit 28 quantizes the low frequency spectrum from the spectrum quantization unit 22, the spectrum characteristic code from the spectrum characteristic determination unit 25, and the quantization extension coefficient or quantization from the extension coefficient quantization unit 27. An envelope coefficient is multiplexed to generate a code string. At this time, the multiplexing unit 28 performs multiplexing after encoding the quantized low-frequency spectrum and the quantization extension coefficient or the quantization envelope coefficient.

As described above, the encoding device 11 determines the spectral characteristics of the input signal based on the low-frequency spectrum feature value and the high-frequency spectrum feature value. Then, the encoding device 11 calculates an expansion coefficient or an envelope coefficient as information for obtaining an extended spectrum at the time of decoding according to the spectrum characteristics.

This makes it possible to obtain an appropriate extended spectrum by using an extension coefficient or an envelope coefficient at the time of decoding, and to obtain high-quality sound even in a low resource environment without increasing the principle delay. In particular, when an extended spectrum is generated using an envelope coefficient, an extended spectrum with a low tonality can be obtained even when a low-frequency spectrum has a high tonality.

<Description of decryption processing>
Next, the decoding process performed by the decoding apparatus 81 when the encoding process described with reference to FIG. 16 is performed by the encoding apparatus 11 will be described with reference to the flowchart of FIG.

In addition, since the process of step S141 and step S142 is the same as the process of step S51 of FIG. 7, and step S52, the description is abbreviate | omitted. However, in step S141, either the quantization expansion coefficient or the quantization envelope coefficient obtained by decomposing the code string is supplied from the decomposition unit 91 to the expansion coefficient inverse quantization unit 93.

In step S143, the extended coefficient inverse quantization unit 93 and the extended spectrum generation unit 94 indicate the tonality in which the low-frequency and high-frequency spectral characteristics are high based on the spectral characteristic code supplied from the decomposing unit 91. Determine whether or not.

For example, when the value of the spectrum characteristic code is “1”, it is determined that the low-frequency and high-frequency spectral characteristics indicate high tonality. In this case, since the code string includes a single quantization extension coefficient, the quantization extension coefficient is supplied from the decomposition unit 91 to the extension coefficient inverse quantization unit 93.

If it is determined in step S143 that the low-frequency and high-frequency spectral characteristics indicate high tonality, the processing in steps S144 and S145 is performed to generate an extended spectrum and supplied to the IMDCT unit 95.

In addition, since the process of these step S144 and step S145 is the same as the process of step S54 of FIG. 7, and step S55, the description is abbreviate | omitted. When the process of step S145 is performed, the process proceeds to step S151.

If it is not determined in step S143 that the low-frequency and high-frequency spectral characteristics indicate high tonality, the process proceeds to step S146.

In step S146, based on the spectral characteristic code, the extended coefficient inverse quantization unit 93 and the extended spectral generation unit 94 indicate tonalities with high spectral characteristics in the low band and low tonal characteristics with high spectral characteristics. Determine whether or not. For example, when the value of the spectrum characteristic code is “2”, it is determined that the low band spectral characteristic indicates a high tonality and the high band spectral characteristic indicates a low tonality.

If it is determined in step S146 that the low band spectral characteristic indicates a high tonality and the high band spectral characteristic indicates a low tonality, the process proceeds to step S147. In this case, the quantization envelope coefficient for each high frequency band is supplied from the decomposition unit 91 to the expansion coefficient inverse quantization unit 93.

In step S147, the expansion coefficient inverse quantization unit 93 inversely quantizes the quantization envelope coefficient for each of a plurality of bands constituting the high frequency band supplied from the decomposition unit 91, and converts the obtained envelope coefficient into an expansion spectrum generation unit. 94. Thereby, for example, the envelope coefficients L31 to L33 of the bands B31 to B33 shown in FIG. 15 are obtained.

In step S148, the extended spectrum generation unit 94 generates an extended spectrum based on the envelope coefficient of each band supplied from the extension coefficient inverse quantization unit 93, and supplies the generated spectrum to the IMDCT unit 95.

Specifically, the extended spectrum generation unit 94 assigns a random number normalized to a value in the range of −1.0 to 1.0 to each frequency bin of the extension band to generate random noise, and the frequency of each band of the random noise The value in the bin is multiplied by the envelope coefficient of each band to obtain an extended spectrum.

After the extended spectrum is generated, the process proceeds to step S151.

Furthermore, if it is not determined in step S146 that the low-frequency spectral characteristics indicate high tonality and the high-frequency spectral characteristics indicate low tonality, processing in steps S149 and S150 is performed.

In this case, the quantization expansion coefficient for each high frequency band is supplied from the decomposition unit 91 to the expansion coefficient inverse quantization unit 93 to perform inverse quantization, and the expansion spectrum is obtained from the expansion coefficient obtained as a result and the low frequency spectrum. Is generated. Note that the processing in step S149 and step S150 is the same as the processing in step S56 and step S57 in FIG.

When the extended spectrum is generated in this way, the process thereafter proceeds to step S151.

When the process of step S145, step S148, or step S150 is performed to generate an extended spectrum, the process of step S151 is performed to generate a time-series signal. The process of step S151 is performed in step S58 of FIG. Since this is the same as the above process, the description thereof is omitted.

When the time-series signal obtained in step S151 is output as an audio signal obtained by decoding, the decoding process ends.

As described above, the decoding device 81 obtains an extension coefficient or an envelope coefficient corresponding to the spectrum characteristic by decoding and inverse quantization, and generates an extension spectrum using the obtained extension coefficient or envelope coefficient.

In this way, the level of the seed spectrum or line dam noise is adjusted using the expansion coefficient or envelope coefficient corresponding to the spectral characteristics to obtain an extended spectrum, whereby the high frequency level can be adjusted in the frequency domain. At the same time, it is possible to achieve high-level level adjustment according to the spectral characteristics. As a result, delay time due to band expansion during decoding can be reduced, and high-quality sound can be obtained even in a low-resource environment.

By the way, the series of processes described above can be executed by hardware or can be executed by software. When a series of processing is executed by software, a program constituting the software is installed in the computer. Here, the computer includes, for example, a general-purpose personal computer capable of executing various functions by installing a computer incorporated in dedicated hardware and various programs.

FIG. 18 is a block diagram illustrating a configuration example of hardware of a computer that executes the above-described series of processes by a program.

In a computer, a CPU (Central Processing Unit) 501, ROM 502, and RAM 503 are connected to each other by a bus 504.

An input / output interface 505 is further connected to the bus 504. An input unit 506, an output unit 507, a recording unit 508, a communication unit 509, and a drive 510 are connected to the input / output interface 505.

The input unit 506 includes a keyboard, a mouse, a microphone, an image sensor, and the like. The output unit 507 includes a display, a speaker, and the like. The recording unit 508 includes a hard disk, a nonvolatile memory, and the like. The communication unit 509 includes a network interface or the like. The drive 510 drives a removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory.

In the computer configured as described above, the CPU 501 loads the program recorded in the recording unit 508 to the RAM 503 via the input / output interface 505 and the bus 504 and executes the program, for example. Is performed.

The program executed by the computer (CPU 501) can be provided by being recorded on the removable medium 511 as a package medium, for example. The program can be provided via a wired or wireless transmission medium such as a local area network, the Internet, or digital satellite broadcasting.

In the computer, the program can be installed in the recording unit 508 via the input / output interface 505 by attaching the removable medium 511 to the drive 510. Further, the program can be received by the communication unit 509 via a wired or wireless transmission medium and installed in the recording unit 508. In addition, the program can be installed in the ROM 502 or the recording unit 508 in advance.

The program executed by the computer may be a program that is processed in time series in the order described in this specification, or in parallel or at a necessary timing such as when a call is made. It may be a program for processing.

The embodiments of the present technology are not limited to the above-described embodiments, and various modifications can be made without departing from the gist of the present technology.

For example, the present technology can take a cloud computing configuration in which one function is shared by a plurality of devices via a network and is jointly processed.

Further, each step described in the above flowchart can be executed by one device or can be shared by a plurality of devices.

Further, when a plurality of processes are included in one step, the plurality of processes included in the one step can be executed by being shared by a plurality of apparatuses in addition to being executed by one apparatus.

Furthermore, the present technology can be configured as follows.

[1]
An acquisition unit for acquiring a low-frequency spectrum and a single expansion coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low-frequency band, or an extension coefficient for each of a plurality of bands constituting the extension band; ,
A generating unit that generates the extended spectrum based on the single extension coefficient or the extension coefficient for each of the plurality of bands;
A decoding device comprising: a combining unit that combines the low-frequency spectrum and the extended spectrum.
[2]
The decoding device according to [1], wherein the generation unit generates the extended spectrum based on the low frequency spectrum and the extension coefficient.
[3]
The decoding device according to [2], wherein the generation unit generates the extended spectrum by adjusting a level of a spectrum obtained from the low frequency spectrum based on the extension coefficient.
[4]
When generating the extension spectrum based on the single extension coefficient, the generation unit adjusts the level of the entire extension band of the spectrum based on the extension coefficient, and the extension for each of the plurality of bands The decoding device according to [3], wherein when generating the extended spectrum based on a coefficient, the level of the band of the spectrum is adjusted based on the extension coefficient of the band.
[5]
The decoding device according to [1], wherein the generation unit generates the extended spectrum by adjusting a predetermined noise level based on the extension coefficient.
[6]
The decoding device according to any one of [1] to [5], wherein the value of the low-frequency spectrum is determined by an amplitude component and a phase component of the original time-series signal.
[7]
The decoding device according to [6], wherein the low frequency spectrum is an MDCT spectrum.
[8]
Obtaining a low band spectrum and a single expansion coefficient for the extension band to obtain an extension spectrum of an extension band different from the low band, or an extension coefficient for each of a plurality of bands constituting the extension band;
Generating the extended spectrum based on the single extension factor or the extension factor for each of the plurality of bands;
A decoding method including a step of combining the low-frequency spectrum and the extended spectrum.
[9]
Obtaining a low band spectrum and a single expansion coefficient for the extension band to obtain an extension spectrum of an extension band different from the low band, or an extension coefficient for each of a plurality of bands constituting the extension band;
Generating the extended spectrum based on the single extension factor or the extension factor for each of the plurality of bands;
A program that causes a computer to execute processing including a step of synthesizing the low-frequency spectrum and the extended spectrum.
[10]
A feature amount extraction unit that extracts a feature amount from a spectrum obtained by orthogonally transforming a time series signal;
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, A calculation unit for calculating based on the spectrum;
An encoding apparatus comprising: a multiplexing unit that multiplexes a low frequency spectrum that is a low frequency component of the spectrum and the extension coefficient to generate a code string.
[11]
The encoding device according to [10], wherein the feature amount is information indicating the tonality of the spectrum.
[12]
The calculation unit calculates the single extension coefficient when the spectrum tonality is high, and calculates the extension coefficient for each of the plurality of bands when the spectrum tonality is low. Encoding device.
[13]
The encoding unit according to any one of [10] to [12], wherein the calculation unit calculates a ratio between an average amplitude of the extension band of the spectrum and an average amplitude of the low band spectrum as the extension coefficient. apparatus.
[14]
The code according to [11], wherein the calculation means calculates envelope information of the extension band of the spectrum as the extension coefficient when the low-frequency tonality of the spectrum is high and the tonality of the extension band of the spectrum is low. Device.
[15]
The encoding device according to any one of [10] to [14], wherein the spectrum value is determined by an amplitude component and a phase component of the time-series signal.
[16]
The encoding device according to [15], wherein the orthogonal transform is MDCT.
[17]
Extract features from the spectrum obtained by orthogonal transform of time series signal,
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, Calculated based on the spectrum,
A coding method including a step of generating a code string by multiplexing a low frequency spectrum which is a low frequency component of the spectrum and the extension coefficient.
[18]
Extract features from the spectrum obtained by orthogonal transform of time series signal,
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, Calculated based on the spectrum,
A program that causes a computer to execute processing including a step of generating a code string by multiplexing a low-frequency spectrum that is a low-frequency component of the spectrum and the extension coefficient.

11 Encoder, 21 MDCT unit, 22 Spectrum quantization unit, 23 Low frequency feature extraction unit, 24 High frequency feature extraction unit, 25 Spectrum characteristic determination unit, 26 Extension coefficient calculation unit, 27 Extension coefficient quantization unit, 28 multiplexing unit, 81 decoding device, 91 decomposing unit, 92 spectrum dequantizing unit, 93 extended coefficient dequantizing unit, 94 extended spectrum generating unit, 95 IMDCT unit

Claims

An acquisition unit for acquiring a low-frequency spectrum and a single expansion coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low-frequency band, or an extension coefficient for each of a plurality of bands constituting the extension band; ,
A generating unit that generates the extended spectrum based on the single extension coefficient or the extension coefficient for each of the plurality of bands;
A decoding device comprising: a combining unit that combines the low-frequency spectrum and the extended spectrum.
The decoding device according to claim 1, wherein the generation unit generates the extended spectrum based on the low frequency spectrum and the extension coefficient.
The decoding device according to claim 2, wherein the generation unit generates the extended spectrum by adjusting a level of a spectrum obtained from the low frequency spectrum based on the extension coefficient.
When generating the extension spectrum based on the single extension coefficient, the generation unit adjusts the level of the entire extension band of the spectrum based on the extension coefficient, and the extension for each of the plurality of bands The decoding device according to claim 3, wherein when generating the extended spectrum based on a coefficient, the level of the band of the spectrum is adjusted based on the extended coefficient of the band.
The decoding device according to claim 1, wherein the generation unit generates the extended spectrum by adjusting a predetermined noise level based on the extension coefficient.
The decoding device according to claim 1, wherein the value of the low-frequency spectrum is determined by an amplitude component and a phase component of the original time-series signal.
The decoding device according to claim 6, wherein the low-frequency spectrum is an MDCT spectrum.
Obtaining a low band spectrum and a single expansion coefficient for the extension band to obtain an extension spectrum of an extension band different from the low band, or an extension coefficient for each of a plurality of bands constituting the extension band;
Generating the extended spectrum based on the single extension factor or the extension factor for each of the plurality of bands;
A decoding method including a step of combining the low-frequency spectrum and the extended spectrum.
Obtaining a low band spectrum and a single expansion coefficient for the extension band to obtain an extension spectrum of an extension band different from the low band, or an extension coefficient for each of a plurality of bands constituting the extension band;
Generating the extended spectrum based on the single extension factor or the extension factor for each of the plurality of bands;
A program that causes a computer to execute processing including a step of synthesizing the low-frequency spectrum and the extended spectrum.
A feature amount extraction unit that extracts a feature amount from a spectrum obtained by orthogonally transforming a time series signal;
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, A calculation unit for calculating based on the spectrum;
An encoding apparatus comprising: a multiplexing unit that multiplexes a low frequency spectrum that is a low frequency component of the spectrum and the extension coefficient to generate a code string.
The encoding device according to claim 10, wherein the feature amount is information indicating the tonality of the spectrum.
The calculation unit according to claim 11, wherein the calculation unit calculates the single extension coefficient when the spectrum tonality is high, and calculates the extension coefficient for each of the plurality of bands when the spectrum tonality is low. Encoding device.
The encoding device according to claim 10, wherein the calculation unit calculates a ratio between an average amplitude of the extension band of the spectrum and an average amplitude of the low band spectrum as the extension coefficient.
The code according to claim 11, wherein the calculation means calculates envelope information of the extension band of the spectrum as the extension coefficient when the tonality of the low band of the spectrum is high and the tonality of the extension band of the spectrum is low. Device.
The encoding device according to claim 10, wherein the spectrum value is determined by an amplitude component and a phase component of the time-series signal.
The encoding apparatus according to claim 15, wherein the orthogonal transform is MDCT.
Extract features from the spectrum obtained by orthogonal transform of time series signal,
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, Calculated based on the spectrum,
A coding method including a step of generating a code string by multiplexing a low frequency spectrum which is a low frequency component of the spectrum and the extension coefficient.
Extract features from the spectrum obtained by orthogonal transform of time series signal,
Depending on the feature amount, a single extension coefficient for the extension band for obtaining an extension spectrum of an extension band different from the low band of the spectrum, or an extension coefficient for each of a plurality of bands constituting the extension band, Calculated based on the spectrum,
A program that causes a computer to execute processing including a step of generating a code string by multiplexing a low-frequency spectrum that is a low-frequency component of the spectrum and the extension coefficient.